Anthropometry, the measurement of human dimensions, is a well-established field with techniques that have been honed over decades of work. The U.S. military, in particular, has performed a number of comprehensive anthropometric studies to provide information for use in the design of military clothing and equipment (Gordon et al., 1989). In recent years, traditional measurement techniques, performed manually with such instruments as calipers and measuring tapes, have been supplemented, and even in some cases replaced, by three-dimensional (3D) scans and digitizers.
THE ANTHROTECH MEASUREMENTS
To establish a comprehensive dataset, Anthrotech investigators used traditional manual measurement techniques to survey 4,026 subjects for 18 facial and head dimensions, as well as height and weight (Box 2-1). Neck circumference was added partway through data collection when it was learned that these metrics are used in the development of some national and international respirator standards (e.g., neck dams for escape hoods) (Anthrotech, 2004).
Before the field study, Anthrotech prepared a measurer’s handbook, which included illustrated instructions for measuring each dimension, as well as a table of values representing allowable measurement error for technicians. Study instruments included an anthropometer, a spreading caliper, a sliding caliper, and a steel measuring tape.
To acquire the anthropometric measurements described above, the investigator must first identify and mark a series of landmarks on the subject’s face (Box 2-2). In many cases this is a relatively straightforward procedure, but there can be complications. For example, many of the landmarks are located by palpating the skin to identify the underlying bone structure. This works well for the traditional measurements done by hand with calipers and measuring tape. However, it can be difficult when using 3D scans (described below) as some bony landmarks are not as readily apparent without palpation.
The accuracy and quality of the anthropometric data collected by Anthrotech is important in that it affects the validity of the respirator fit-test face panels derived from these data, which ultimately impact respirator face panels and certification. In general, data validity could be described by accuracy—a measure of how close the measurements are to an evidence-based best practice standard or gold standard if there is one—and precision (the reliability, repeatability, reproducibility, and consistency of measurements of the same person). Measurement errors include errors in both accuracy and precision; measurements from the same person can be both inaccurate, imprecise, both, or neither. Measurement errors may arise from: (1) human error, which in this case could include inaccurate and imprecise identification and placement of landmarks and dimension measurements from field technicians, in addition to other human errors such as data entry errors; (2) equipment error (less a
problem in the NIOSH-sponsored Anthrotech study as the calipers and measuring tapes are standard equipment that are calibrated and have little chance of change); (3) systematic bias; and (4) random error. Human measurement error in this type of study often is a major source of error.
Ideally, in a large field study on anthropometric measurement, a pilot study is conducted to: (1) determine the accuracy of the placement of landmarks; (2) determine the accuracy of measuring distances between marked landmarks (these would both be done to evaluate intraobserver error); and (3) determine the contribution of multiple observers to measurement error (interobserver error). This allows the assessment of how much error can be expected in the study and determine the sources of error, thereby offering a means for improvement. For example, if interobserver error is greater than intraobserver error, technicians can identify ways to improve the accuracy of their measures, which should decrease both inter- and intraobserver error.
The NIOSH-sponsored Anthrotech study made special efforts to reduce human error, but did not adequately account for potential error in measurements. Their quality control measures included providing a handbook with detailed measurement instructions and having the field technicians practice on each other until an allowable level of accuracy was achieved in identifying and marking certain landmarks and dimension measurements. The accuracy was controlled to be between 1 and 3 mm, depending on the dimension measured (Zhuang and Bradtmiller, 2005; Anthrotech, 2004), which was based on standard practice in the anthropometric field (Gordon et al., 1989). However, the committee did not find data in the 2004 report and all the materials it reviewed that could be used to assess the precision of measurements from the technicians (intraobserver and interobserver variations in measurements) (Anthrotech, 2004). In addition, the committee did not find any record in the protocol, report, or other materials it reviewed of an attempt to evaluate the potential relation between biological variation (individual differences in facial shape) and measured error. A useful measure in this regard can be found in the panel report of the 1988 Army survey of anthropometric measures (Gordon et al., 1989). More examples of designs for estimating measurement error are available from such articles as Aldridge et al. (2005), Corner and Richtsmeier (1992), Kohn and J. (1992), Richtsmeier et al. (1995), Valeri et al. (1998), Weinberg and Kolar (2005), and Weinberg et al. (2005).
The NIOSH-sponsored Anthrotech report did not adequately address the potential impact of measurement error on the validity and quality of the anthropometric face dimension data.
Recommendation 2-1: Analyze Measurement Error.
In future studies NIOSH should perform additional analyses of the impact of measurement error, including the effects of intraobserver and interobserver variations in measurement.
To develop a set of standard head forms, NIOSH requested that Anthrotech perform 3D scans on approximately a quarter of the subjects (1,045 individuals).1 However, NIOSH did not regard the 3D scans as a potential source of data for face panel revision. Given the perceived limitation of 3D scans and the investigators’ concern over the time and expense required to relocate the scanner to each site, the investigators chose to scan only a portion of the subjects (Anthrotech, 2004). The datasets from these 3D scans were archived. NIOSH informed the committee that the decision not to use the 3D scan data was due to discrepancies between the 3D scan data and the traditional measures (Table 2-1). Although the causes of the differences are unknown, many of these are well outside the 1 to 3mm acceptable margin of error. NIOSH did not explain to the committee how it was decided that the traditional data were more accurate or more reliable than the 3D data. The 2004 report does reference that absence of standardized data analysis methods (Anthrotech, 2004); however, this in itself does not explain why the traditional data
TABLE 2-1 Differences Between Traditional and 3D Measurements
may be more reliable than the 3D data. The study published by Anthrotech indicates that, upon review of the 3D measurements, there were no landmarking errors, thus suggesting that any discrepancies may have resulted from errors in manual measurements (Anthrotech, 2004).
The decision to focus solely on the traditional data and to ignore the 3D scan data is of concern because the 3D scan data have several advantages over traditional manually collected data, including the following:
3D scan data generally have better resolution than caliper data and provide a more accurate and complete summary of the facial geometry.
3D scan data can provide information on the localized variation around a landmark, which, if applied correctly, can be useful for designing masks.
Databases from 3D scans may be reanalyzed if additional measurements are required.
3D scan data often rely on non-bony landmarks (Valeri et al., 1998), which eliminates the need for palpating the bony landmarks during traditional manual measurements.
Collecting measures from an image is a relatively simple process with the advantage of checking measurements frequently.
Surface data may be used in mathematical modeling and simulation analysis (finite element analysis and animation).
If the leaks in respirator fit testing can be identified at certain parts of the respirator seal, 3D contour data can be reanalyzed to further correlate local features around the leak with the fit factor of the mask.
Data from the 3D scanned images can be used to derive all of the traditional anthropometric measures, such as linear distances and angles, but they can also be used to derive a variety of new measures, such as surface contours or arcs along surfaces, and one or more of these new measures might offer ways to improve fit that are not possible with the traditional data. Of course, working with such novel measures may require the development of new analytical tools of greater complexity, but given the advantages of the 3D data, it seems to be a direction worth pursuing.
Validation of Three-Dimensional Scan Data
As described above there are many potential advantages of 3D scan data over traditionally collected anthropometric data. However, before 3D surface scans can be used, further data validation must be conducted. These efforts will require establishing an evidence-based best practice standard against which measures collected from the 3D image data can be compared. NIOSH should validate the 3D scans against an external gold standard already proven to be accurate. One potential method would be to use point digitizers to acquire the same landmarks or linear measures as collected from the 3D scan on a mannequin head where the measurements are already known. This would allow repeat measures for validation and verification of the 3D scan data collected from scans. Mannequin tests should be followed by tests of live individuals, so that the effects of movement and possibly day-to-day variation (e.g., changes in hydration, or possibly from injury) may be evaluated.
A study of the precision of data collected from 3D scan images will require comparisons of relevant data sets at several levels, including (1) a comparison of multiple datasets collected from single images; (2) a comparison of datasets collected from multiple images of the same subject; and (3) a comparison of various datasets collected from several images of several specimens or subjects.
Three-dimensional scan data may offer advantages over traditional, manually-collected anthropometric data; however, there is no evidence base of best practice against which 3D scans may be compared.
Recommendation 2-2: Consider Utilizing Three-Dimensional Scan Data.
NIOSH should consider collecting and utilizing data from 3D scans, either alone, or in combination with traditional manually collected data, to ensure the most robust set of data are used to develop future anthropometric face panels.
Aldridge, K., S. A. Boyadjiev, G. T. Capone, V. B. DeLeon, and J. T. Richtsmeier. 2005. Precision and error of three-dimensional phenotypic measures acquired from 3dMD photogrammetric images. Am J Med Genet A 138(3):247-253.
Anthrotech. 2004. A head-and-face anthropometric survey of U.S. respirator users: Final report. Prepared by B. Bradtmiller and M. Friess for NIOSH/NPPTL.
Corner, B. D., and J. T. Richtsmeier. 1992. Cranial growth in the squirrel monkey (saimiri sciureus): A quantitative analysis using three-dimensional coordinate data. Am J Phys Anthropol 87(1):67-81.
Gordon, C. C., T. Churchill, C. E. Clauser, B. Bradtmiller, J. T. McConville, I. Tebbetts, and R. A. Walker. 1989. 1988 anthropometric survey of U.S. Army personnel: Methods and summary statistics. United States Army Natick research, development, and engineering center.
Kohn, L., and Cheverud, J. 1992. Calibration, validation, and evaluation of scanning systems: Anthropometric imaging system repeatability, electronic imaging of the human body. Paper presented at Proceedings of a working group: CSERIAC, Dayton, OH. Pp. 114-123.
Richtsmeier, J. T., C. H. Paik, P. C. Elfert, T. M. Cole, 3rd, and H. R. Dahlman. 1995. Precision, repeatability, and validation of the localization of cranial landmarks using computed tomography scans. Cleft Palate Craniofac J 32(3):217-227.
Valeri, C. J., T. M. Cole, 3rd, S. Lele, and J. T. Richtsmeier. 1998. Capturing data from three-dimensional surfaces using fuzzy landmarks. Am J Phys Anthropol 107(1):113-124.
Weinberg, S. M., and J. C. Kolar. 2005. Three-dimensional surface imaging: Limitations and considerations from the anthropometric perspective. J Craniofac Surg 16(5):847-851.
Weinberg, S. M., N. M. Scott, K. Neiswanger, and M. L. Marazita. 2005. Intraobserver error associated with measurements of the hand. Am J Hum Biol 17(3):368-371.
Zhuang, Z., and B. Bradtmiller. 2005. Head-and-face anthropometric survey of U.S. respirator users. J Occup Environ Hyg 2(11):567-576.